Abstract: Aspects for forward propagation of a convolutional artificial neural network are described herein. The aspects may include a direct memory access unit configured to receive input data from a storage device and a master computation module configured to select one or more portions of the input data based on a predetermined convolution window. Further, the aspects may include one or more slave computation modules respectively configured to convolute a convolution kernel with one of the one or more portions of the input data to generate a slave output value. Further still, the aspects may include an interconnection unit configured to combine the one or more slave output values into one or more intermediate result vectors, wherein the master computation module is further configured to merge the one or more intermediate result vectors into a merged intermediate vector.

Abstract: Aspects of the disclosure generally relate to computing devices and may be generally directed to devices, systems, methods, and/or applications for learning conversations among two or more conversation participants, storing this knowledge in a knowledgebase (i.e. neural network, graph, sequences, etc.), and enabling a user to simulate a conversation with an artificially intelligent conversation participant.

Abstract: An approach is provided in which a knowledge manager identifies a first cohort type and a second cohort type corresponding to an entity included in a question. The knowledge manager determines inferred states to the question by comparing a first set of cohort attributes corresponding to a first cohort type against entity attributes corresponding to the question. In turn, the knowledge manager generates possible answers by comparing the inferred states against a second set of cohort attributes corresponding to a second cohort type.

Abstract: A predictive modeling method may include obtaining a fitted, first-order predictive model configured to predict values of output variables based on values of first input variables; and performing a second-order modeling procedure on the fitted, first-order model, which may include: generating input data including observations including observed values of second input variables and predicted values of the output variables; generating training data and testing data from the input data; generating a fitted second-order model of the fitted first-order model by fitting a second-order model to the training data; and testing the fitted, second-order model of the first-order model on the testing data. Each observation of the input data may be generated by (1) obtaining observed values of the second input variables, and (2) applying the first-order predictive model to corresponding observed values of the first input variables to generate the predicted values of the output variables.

Abstract: Methods, systems and computer program products are provided for cross-media recommendation by store a plurality of taste profiles corresponding to a first domain and a plurality of media item vectors corresponding to a second domain. An evaluation taste profile in the first domain is applied to a plurality of models that have been generated based on relationship among the plurality of taste profiles and the plurality of media item vectors, and obtain a plurality of resulting codes corresponding to at least one of the plurality of media item vectors in the second domain.

Abstract: Provided herein are systems, methods and computer readable media for classifying a provider of products, services or experiences as a provider that should be engaged based on a predicted return rate for any products, services or experiences that may be offered and purchased by a consumer. An example method may comprise supplying a classifying model with a dataset, wherein the dataset comprises an identification of a provider and a plurality of attributes corresponding to the provider and identifying a class of the provider in accordance with the plurality of corresponding attributes, wherein the identification is determined based on one or more patterns determinative of a return rate by the classifying model.

Abstract: An information processing apparatus including inter-class node insertion means for inserting an input vector into a network as an inter-class insertion node. The apparatus further includes a winner node learning time calculation means for incrementing, when an edge is connected between a first winner node and a second winner node, learning time of a node for the first winner node by a predetermined value.

Abstract: This disclosure relates to a method for estimating a particle size distribution (PSD) and a morphology for a set of particles. A computer system receives a plurality of chord length distributions (CLDs) of different types for a set of particles. The computer system then estimates a morphology for the set of particles based on the plurality of received CLDs. The computer system also identifies a plurality of descriptors of the plurality of CLDs for the set of particles based on the plurality of received CLDs. The computer system then estimates metrics for the PSD for the set of particles based on the plurality of identified CLD descriptors. Based on the estimated PSD metrics for the set of particles, the computer system generates an estimate of the PSD for the set of particles. Finally, the computer system outputs the estimated morphology and the estimated PSD for the set of particles.

Abstract: Technical solutions are described to accelerate training of a multi-layer convolutional neural network. According to one aspect, a computer implemented method is described. A convolutional layer includes input maps, convolutional kernels, and output maps. The method includes a forward pass, a backward pass, and an update pass that each include convolution calculations. The described method performs the convolutional operations involved in the forward, the backward, and the update passes based on a first, a second, and a third perforation map respectively. The perforation maps are stochastically generated, and distinct from each other. The method further includes interpolating results of the selective convolution operations to obtain remaining results. The method includes iteratively repeating the forward pass, the backward pass, and the update pass until the convolutional neural network is trained. Other aspects such as a system, apparatus, and computer program product are also described.

Abstract: Methods, systems, and apparatus, including computer programs encoded on computer storage media, for receiving a request from a client to process a computational graph; obtaining data representing the computational graph, the computational graph comprising a plurality of nodes and directed edges, wherein each node represents a respective operation, wherein each directed edge connects a respective first node to a respective second node that represents an operation that receives, as input, an output of an operation represented by the respective first node; identifying a plurality of available devices for performing the requested operation; partitioning the computational graph into a plurality of subgraphs, each subgraph comprising one or more nodes in the computational graph; and assigning, for each subgraph, the operations represented by the one or more nodes in the subgraph to a respective available device in the plurality of available devices for operation.

Abstract: Disclosed herein is a technique for implementing a framework that enables application developers to enhance their applications with dynamic adjustment capabilities. Specifically, the framework, when utilized by an application on a mobile computing device that implements the framework, can enable the application to establish predictive models that can be used to identify meaningful behavioral patterns of an individual who uses the application. In turn, the predictive models can be used to preempt the individual's actions and provide an enhanced overall user experience. The framework is configured to interface with other software entities on the mobile computing device that conduct various analyses to identify appropriate times for the application to manage and update its predictive models. Such appropriate times can include, for example, identified periods of time where the individual is not operating the mobile computing device, as well as recognized conditions where power consumption is not a concern.

Abstract: The present disclosure describes the use of growth models and data driven models that are combined for quickly and efficiently modeling SAGD reservoir oil production. Growth function surrogate models are used for efficient and reliable reservoir modeling and production forecasting as opposed to CPU intensive simulations based on finite difference models. A data-driven technique can then compare the growth function surrogate model with real field data to find discrepancies and inconsistencies between the two, allowing for an updates and improvements of the growth function model.

Abstract: Providing predictive data predicting data values for a historical dataset. The method facilitates improving the accuracy of the predictive data by identifying for a user, and allowing the user to select ancillary datasets that can be evaluated, using a predictive evaluation, together with a historical dataset to improve the accuracy of the predictive data. A user interface is provided to a user. The user interface identifies one or more ancillary datasets. The ancillary datasets are identified to the user based on the ancillary datasets meeting a threshold condition to a historical dataset. The ancillary datasets are selectable by the user in the user interface. User input is received at the user interface selecting one or more of the ancillary datasets. A predictive dataset is displayed to the user. The predictive dataset is determined by predictive evaluation of the historical dataset and the one or more selected ancillary datasets.

Abstract: A mobile device includes a frontend application, a prediction layer including a dispatch unit, prediction generation unit, metadata store, and curve fitting unit. A method includes receiving at the prediction layer a frontend application service request, forwarding the service request about contemporaneously to the curve fitting unit, prediction generation unit, and a backend server, the prediction generation unit searching the metadata store for a predictive formula associated with the service request, calculating a response using the predictive formula, and providing the calculated response for display in a user interface as an interim result to the service request. A response from the backend server is displayed by the frontend application. The curve fitting unit generates and/or refines a predictive formula for the service request based on the service request parameters and the backend server response. A non-transitory computer-readable medium is also disclosed.

Abstract: One embodiment of an accelerator includes a computing unit; a first memory bank for storing input activations and a second memory bank for storing parameters used in performing computations, the second memory bank configured to store a sufficient amount of the neural network parameters on the computing unit to allow for latency below a specified level with throughput above a specified level. The computing unit includes at least one cell comprising at least one multiply accumulate (“MAC”) operator that receives parameters from the second memory bank and performs computations. The computing unit further includes a first traversal unit that provides a control signal to the first memory bank to cause an input activation to be provided to a data bus accessible by the MAC operator. The computing unit performs computations associated with at least one element of a data array, the one or more computations performed by the MAC operator.

Abstract: A predictive modeling method may include determining a time interval of time-series data; identifying one or more variables of the data as targets; determining a forecast range and a skip range associated with a prediction problem represented by the data; generating training data and testing data from the time-series data; fitting a predictive model to the training data; and testing the fitted model on the testing data. The forecast range may indicate a duration of a period for which values of the targets are to be predicted. The skip range may indicate a temporal lag between the time period corresponding to the data used to make predictions and the time period corresponding to the predictions. The skip range may separate input data subsets representing model inputs from subsets representing model outputs, and separate test data subsets representing model inputs from subsets representing validation data.

Abstract: In some scenarios, devices may execute applications that are configured to monitor a set of conditions (e.g., geographic coordinates detected by global positioning system (GPS) receivers), and to execute actions upon detecting the fulfillment of the conditions. However, in such architectures, each application may be responsible for polling the sensors of the device to detect condition fulfillment, and it may be difficult to specify rules in a hardware-independent manner involving multiple applications and/or devices. Presented herein are techniques for configuring devices to perform actions by receiving a rule set from a rule server; registering a set of condition tests for respective conditions of a rule with the sensors of the device; upon being notified by the sensor that a condition test has been fulfilled, evaluating the conditions of the rule; and upon determining a fulfillment of the conditions, executing one or more actions (optionally involving multiple devices and/or applications).

Abstract: A method for applying personalized machine learning models is provided. The method includes producing one or more feature vectors that represents features of one of a plurality of files of a file system and selecting, from a plurality of personalized machine learning models that model user accesses to the files of the file system a subset of the personalized machine learning models each of which has a plurality of non-zero weights corresponding to non-zero features of the one or more feature vectors. The method includes determining from the subset of personalized machine learning models which users of a plurality of users of the file system are likely to access the one of the plurality of files.

Abstract: A method is provided for processing on an acceleration component a machine learning classification model. The machine learning classification model includes a plurality of decision trees, the decision trees including a first amount of decision tree data. The acceleration component includes an acceleration component die and a memory stack disposed in an integrated circuit package. The memory die includes an acceleration component memory having a second amount of memory less than the first amount of decision tree data. The memory stack includes a memory bandwidth greater than about 50 GB/sec and a power efficiency of greater than about 20 MB/sec/mW.

Abstract: A content rewriting system is described herein that allows web site administrators to setup rewriting of web responses in an easy and efficient manner. The system provides a configuration schema and an efficient workflow that enables web administrators to easily setup rules to modify HTML or other content without having a high performance penalty or losing flexibility. The content rewriting system applies regular expressions or wildcard patterns to a response to locate and replace the content parts based on the rewriting logic expressed by outbound rewrite rules. The system parses an initial response generated by a web application, applies one or more outbound rules to rewrite the response, and provides the rewritten response to a client that submitted a request for the response.